Hi Stéfan 3D skeletonization would be a great idea. I wonder if the algorithm you mention is the same as the one used in http://fiji.sc/Skeletonize3D (which is quite popular in the X-ray tomography community). There are so many 3D skeletonization algorithms that understanding which specs are required might be an important first step. Can you explain which kind of sponsoring it would be? Is it only available for people living in the US, or in other countries? For students only? Cheers Emma On Mon, Nov 02, 2015 at 02:26:46PM -0800, Stefan van der Walt wrote:

Hi all,

I have been approached by a group that is interested in sponsoring the development of 3D skeletonization in scikit-image. One potential starting place would be:

http://www.insight-journal.org/browse/publication/181

Is anyone interested in working on this? Please get in touch.

Thanks! Stéfan

Should we use/apply this to a particular volumetric dataset while prototyping different methods, to ensure accurate comparisons? Should anisotropic, regularly sampled voxels be supported?

I suggest the lobster, one of the bonsai, or the XMas tree datasets located here as excellent 'torture tests' for 3d skeletonization. http://www9.informatik.uni-erlangen.de/External/vollib/

Hi, I don’t want to volunteer for this project, but I just wanted to mention that the 3D skeletonization from ITK is easily accessible to Python through SimpleITK, see example below for the lobster dataset. SimpleITK could be used for comparison or validation of the proposed scikit-image algorithm. Kind Regards, Kevin PS: is there another way to load those *.pvm datasets in Python without converting them to raw and hardcoding the image dimension and pixel type? An skimage.io.imread() plugin?

On 3 Nov 2015, at 21:18, Emmanuelle Gouillart wrote:

...

Don't forget that Fiji's code is mostly GPL, so don't try to copy it, at least not without first discussing dual licensing with the author(s).

Excellent point. Actually I'm only using the Fiji page as a way to find the paper by Lee et al. again :-).

...

On Tue, Nov 3, 2015 at 12:55 PM, Josh Warner wrote:

...

Should we use/apply this to a particular volumetric dataset while prototyping different methods, to ensure accurate comparisons?

...

Should anisotropic, regularly sampled voxels be supported?

-- You received this message because you are subscribed to the Google Groups "scikit-image" group. To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe@googlegroups.com. For more options, visit https://groups.google.com/d/optout.

It looks like the lobster and bonsai can be downloaded directly as raw volumes (8 bit only, but will serve these purposes perfectly well) here: http://www.volvis.org/ This simple wrapper for np.fromfile will load them import numpy as np def loadraw(rawfile, shape=None, dtype=np.uint8): """ Load RAW volume to a NumPy array. Parameters ---------- rawfile : string Path to *.raw volume. shape : tuple Shape of the volume. If not provided, output will be a rank-1 stream which can be reshaped as desired. dtype : NumPy dtype Dtype of the raw image volume. """ vol = np.fromfile(rawfile, dtype=dtype) if shape is not None: vol = vol.reshape(shape) return vol For the lobster, use shape=(56, 324, 301) and recall the voxel spacing has a ratio of 1.4:1:1 For the bonsai, use shape=(256, 256, 256) and the volume is isotropic (1:1:1 spacing) On Monday, November 9, 2015 at 8:42:26 PM UTC-5, stefanv wrote: Hi Kevin

On 2015-11-07 09:46:17, 'Kevin Keraudren' via scikit-image < scikit-image@googlegroups.com> wrote:

...

I don’t want to volunteer for this project, but I just wanted to mention that the 3D skeletonization from ITK is easily accessible to Python through SimpleITK, see example below for the lobster dataset. SimpleITK could be used for comparison or validation of the proposed scikit-image algorithm.

Thanks for the pointer. In this case, one of the purposes of the exercise is to stay away from a heavy dependency such as ITK.

...

PS: is there another way to load those *.pvm datasets in Python without converting them to raw and hardcoding the image dimension and pixel type? An skimage.io.imread() plugin?

I have no idea about .pvm files, but perhaps we should start a set of plugin gists on the wiki somewhere?

Stéfan

​